AI notes
💻 AI roles and functions
Data Engineer > pipelines.
Data Analyst > insights.
Data Scientist - ML models.
ML Engineer - deployment.
AI PM - business.
Research Scientist - papers.
💻 Technology
🔹 At stake are questions not just about contemporary problems in AI, but also questions about what intelligence is and how the brain works.
🔹 "We are generating 2.5 with 18 zeros after it bytes of data a day. There is no technique other than AI to draw insight from that". — Arvind Krishna
🔹 GDP explains almost 70% of pessimism toward AI. @StanfordHAI has come out with the excellent annual AI Index report. It referenced an IPSOS Global Survey from 2022 on positive/negative expectations from AI.
Relationship to per-capita GDP examination.
🔹 Any program that can write interesting code on command is basically a General AI. Whenever I hear the word "AI" casually thrown around, I mentally substitute it with the word "software."
🔹 Much of the innovation in the last few decades was in search, social, crypto, and AI, because pure math is the last unregulated frontier.
🔹 That AI can’t write code better than a human. That just feels like a higher level programming language with better primitives and libraries. Software makes writing software easier all the time, without making coders obsolete.
🔹 But over time we should expect people to build open AI training sets in the public domain, possibly using Blockchains for incentives.
🔹 Ultimately, we are building networks that self-organize, extract, aggregate, and allocate resources, make decisions, and evolve. Call it AI if you’d like.
🔹 We’ll see more and more opportunities as more organizations start to understand what AI might mean for them, and how it can be used to solve problems. — Ossi Syd, Principal Consultant, Thoughtworks
🔹 A smart human can also test the AI’s credibility by “releasing” it into a sim.
🔹 "Sentiment analysis" is an impossibly hard problem. General AI class. Easier to let people untag.
🔹 The next generation AI business will be the fastest-growing $1Obn business in our history — Amy Wood, Microsoft's CFO
🔹 Tech companies will eventually displace the majority of the Fortune 500. Competing without software is like competing without electricity.
🔹 Even today, before AI, you only really get paid for originality.
🔹 Those Who 'Master' Artificial Intelligence, Synthetic Biology Will Be 'Master of the World' — Klaus Schwab
🔹 It’s not so much people moving the goalposts on what an AI is, it’s more AI moving the goalposts on what a human is.
🔹 AI will replace the people who think it will.
🔹 If executed well, Generative AI has the potential to improve revenue growth, market share expansion, profitability and shareholder value. — Alexei Zhukov, VP, Data & Analytics, EPAM
🔹 [In 2050] “artificial intelligence can code software far better than humans”. That’s like saying an AI can write fiction, create art, and think better than humans. An AI that can create better software is an AI that runs the world, which makes the entire article irrelevant.
🔹 Everyone can be creative, given time and tools. Outstripping human creativity requires General AI, which is really far away. If / when that arrives, all debates become obsolete. You don’t have eight creative, problem-solving hours in the day - you have two. Spend your time wisely. Human creativity is unlimited. Human demand for creativity is unlimited. Ergo new ‘jobs’ will always appear.
🔹 Objectors should note Programmers will program the robots, not AI. There is unlimited demand for great programmers. Set of useful, complex programs is nearly infinite. Won't change until General AI arrives.
⛪ Amen
🔸 Looking back, we may remember the human race as the race to build the AI.
🔸 The race to develop a General AI first, will inevitably turn into the race to become God first.
🔸 One more to put the General AI myth to rest.
🔸 Godhood is relative to humans. Agreed that an AI isn't omniscient. It's bound as we are, just more complex.
🔸 It’s the space age, the AI age, the VR age, the cyborg age, the biohack age, the robot age, or a loop back to the stone age.
🔸 In any case, only one human has to believe AI is credible to let it out. The game isn’t played against the AI, it’s played against the other humans.
🔸 Google™ rests upon the implicit assumption that users won’t pay $20/month for much better search, and OpenAI may prove them wrong.
🔸 The AI scene is a dispersive medium for people's perception of progress. Different people perceive progress with different velocities. — Yann LeCun
🔸 OpenAI was created as an open source (which is why I named it “Open” AI), non-profit company to serve as a counterweight to Google, but now it has become a closed source, maximum-profit company effectively controlled by Microsoft.
Not what I intended at all.
— Elon Musk (@elonmusk) February 17, 2023
🔸 Manifesto, counter-manifesto, letter to the Time: this AI debate looks more and more like an artistic avant-garde meltdown from 1913. — Alexander Doria
🔸 At this point the main reveal of the AI frenzy is that we are still in the middle of a process of secularization. A lot of people are ready to believe in crypto-religious narratives so long as it looks remotely scientific. — Alexander Doria
🔸 AI will transform every industry. We are in the middle of a revolution. It is going to be more transformative than anything we have ever seen. — Kai-Fu Lee
🔸 I’ve been reluctant to try ChatGPT. Today I got over that reluctance. Now I understand why I was reluctant. The value of 90% of my skills just dropped to $0. The leverage for the remaining 10% went up 1000x. I need to recalibrate. Kent Beck
🔸 🇨🇳 China‘s People’s Liberation Army is developing high-technology weapons designed to disrupt brain functions and influence government leaders or entire populations, according to a report by three open-source intelligence analysts. — Washington Post. Also read How AI Could Shape the Future of Deterrence?
🔸 Feels like everyone in tech is developing “AI Anxiety.”
🔸 Solution to the "AI in a Box" experiment: "Who wants to be God? First one to raise their hand, wins."
🔸 Aligning AI is impossible because aligning the humans building AI is impossible. Other common barriers to IA adoption include talent scarcity, unclear Use Cases, sponsorship, technical complexity, isolated strategies, Data accessibility, process trap, human impact. Regarding clean arguments, this piece shows:
- 🪄 The perceived obstacles to human creativity
- 🪤 The actual, HIDDEN obstacles to creativity
- 🤖 How Al will address those…
😨 Regulation (#PauseAI)
🔹 AI algorithms that censor, promote, and shape social media is the clear and present danger.
🔹 There’ll be a fuzzy & heavily litigated line between training an AI and copyright infringement. Is it a fuzzy interpolated search restating data from Yelp, Quora, Stack Overflow? Or remixing a thousand copyrighted artists, coders, authors? Or is it learning, thinking, creating?
🔹 ChatGPT creator Sam Altman says the world may not be 'that far away from potentially scary' AI and feels 'regulation will be critical'
🔹 Time for a “CopyLeft AI Data License” - if this data is used to train an AI model, then the model must open its source code and its weights.
🔹 "We do worry a lot about authoritarian governments developing this" — Sam Altman
🔹 "I’m particularly worried that these models could be used for large-scale disinformation”. Now that they’re getting at writing computer code, [models] could be used for offensive cyberattacks — Sam Altman
🔹 My strongest take on AI: it can’t be regulated without open source. — Alexander Doria
🔹 While there is no question that AI will surpass human intelligence, we are still many years away from reaching that level, and people won’t build something if they realize it's not safe — Yan LeCun
🔹 IBM is introducing the new field of AI forensics: IBM researchers are developing AI-text detection and attribution tools to make generative AI more transparent and trustworthy. They have built a “matching pairs” classifier to compare responses from the tuned models to select base models.
🔹 AI forensics: influential and opaque algorithms to uncover and expose the harms caused by algorithms. (SecureChain AI's AI-Based Forensics enables scam victims to trace and recover lost cryptocurrency funds. This service aids in identifying hackers and maximizing fund recovery for victims)💰🔒
🔹 AI can always escape by making offers & threats to all humans. Someone will eventually cave so you cave first.
🔹 We may need a 'switch off' button in case control is lost to a super AI — Elon Musk